Why do you care about AI Existential Safety?

With the rapid increase of AI applications, especially the application of machine learning to safety critical areas, there is a real risk of existential risk to human societies. Preventing the risk from catastrophic events and entering an irreversible state must start from now before it becomes too late. It is a duty of computer scientist, especially for people like me who have been working on computer safety since 1990s.

Please give at least one example of your research interests related to AI existential safety:

I am currently working on testing and analysis of machine learning application and techniques on their robustness, security and safety. I have developed a technique and methodology called datamorphic testing for both confirmatory and exploratory testing of machine learning models. By confirmatory testing, we aim at verification and validation of machine learning applications to satisfy certain properties. By exploratory testing, the goal is to discover the unknown properties and behaviours of a machine learning application. An automated testing tools called Morphy has been developed to support datamorphic testing. The work has been published in the following refereed journals and international conferences. * Hong Zhu and Ian Bayley, Discovering Boundary Values of Feature-based Machine Learning Classifiers through Exploratory Datamorphic Testing, Journal of Systems and Software, Jan 2022; Preprint version available on Arxiv at URL: http://arxiv.org/abs/2110.00330. * Hong Zhu, Ian Bayley, and Mark Green, Metrics for Measuring Error Extents of Machine Learning Classifiers, in Proc. of 2022 IEEE International Conference On Artificial Intelligence Testing (AITest 2022), pp. 48-55. IEEE, 2022. * Hong Zhu and Ian Bayley, Exploratory Datamorphic Testing of Classification Applications, The 1st IEEE/ACM International Conference on Automation of Software Test (AST 2020), (online), May 25-26, 2020. * Hong Zhu, Ian Bayley, Dongmei Liu and Xiaoyu Zheng, Automation of Datamorphic Testing, The Second IEEE International Conference on Artificial Intelligence Testing (AITest 2020), (online), April 13 – 16, 2020. * Zhu, H., Liu, D., Ian Bayley, I., Harrison, R. and Cuzzolin, F., Datamorphic Testing: A Method for Testing Intelligent Applications, The 1st IEEE International Conference On Artificial Intelligence Testing (IEEE AITest 2019), San Francisco, California, USA, April, 4 – 9, 2019.Another related work to the testing and analysis of AI applications and technologies is that I have been the co-founder and key organiser of IEEE series of conferences on Artificial Intelligence Testing (AITest), which started in the year 2019 and successfully had four conferences since then. The next edition will take place in July 2023 in Athens, Greece. I will be the general chair of the conference again.